1 /*
2 * SPDX-FileCopyrightText: Copyright OpenBMC Authors
3 * SPDX-License-Identifier: Apache-2.0
4 */
5
6 #include "NvidiaGpuPowerSensor.hpp"
7
8 #include "MctpRequester.hpp"
9 #include "NvidiaSensorUtils.hpp"
10 #include "SensorPaths.hpp"
11 #include "Thresholds.hpp"
12 #include "Utils.hpp"
13 #include "sensor.hpp"
14
15 #include <bits/basic_string.h>
16
17 #include <NvidiaDeviceDiscovery.hpp>
18 #include <NvidiaGpuMctpVdm.hpp>
19 #include <OcpMctpVdm.hpp>
20 #include <phosphor-logging/lg2.hpp>
21 #include <sdbusplus/asio/connection.hpp>
22 #include <sdbusplus/asio/object_server.hpp>
23
24 #include <cstddef>
25 #include <cstdint>
26 #include <functional>
27 #include <limits>
28 #include <memory>
29 #include <optional>
30 #include <span>
31 #include <string>
32 #include <system_error>
33 #include <utility>
34 #include <vector>
35
36 using namespace std::literals;
37
38 static constexpr double gpuPowerSensorMaxReading = 5000;
39 static constexpr double gpuPowerSensorMinReading =
40 std::numeric_limits<uint32_t>::min();
41
NvidiaGpuPowerSensor(std::shared_ptr<sdbusplus::asio::connection> & conn,mctp::MctpRequester & mctpRequester,const std::string & name,const std::string & sensorConfiguration,uint8_t eid,uint8_t sensorId,sdbusplus::asio::object_server & objectServer,std::vector<thresholds::Threshold> && thresholdData,const gpu::DeviceIdentification deviceType)42 NvidiaGpuPowerSensor::NvidiaGpuPowerSensor(
43 std::shared_ptr<sdbusplus::asio::connection>& conn,
44 mctp::MctpRequester& mctpRequester, const std::string& name,
45 const std::string& sensorConfiguration, uint8_t eid, uint8_t sensorId,
46 sdbusplus::asio::object_server& objectServer,
47 std::vector<thresholds::Threshold>&& thresholdData,
48 const gpu::DeviceIdentification deviceType) :
49 Sensor(escapeName(name), std::move(thresholdData), sensorConfiguration,
50 "power", false, true, gpuPowerSensorMaxReading,
51 gpuPowerSensorMinReading, conn),
52 eid(eid), sensorId{sensorId},
53
54 mctpRequester(mctpRequester), objectServer(objectServer)
55
56 {
57 std::string dbusPath = sensorPathPrefix + "power/"s + escapeName(name);
58
59 sensorInterface = objectServer.add_interface(
60 dbusPath, "xyz.openbmc_project.Sensor.Value");
61
62 for (const auto& threshold : thresholds)
63 {
64 std::string interface = thresholds::getInterface(threshold.level);
65 thresholdInterfaces[static_cast<size_t>(threshold.level)] =
66 objectServer.add_interface(dbusPath, interface);
67 }
68
69 association = objectServer.add_interface(dbusPath, association::interface);
70
71 setInitialProperties(sensor_paths::unitWatts);
72
73 const std::optional<std::string> physicalContext =
74 nvidia_sensor_utils::deviceTypeToPhysicalContext(deviceType);
75
76 if (physicalContext)
77 {
78 commonPhysicalContextInterface = objectServer.add_interface(
79 dbusPath, "xyz.openbmc_project.Common.PhysicalContext");
80
81 commonPhysicalContextInterface->register_property("Type",
82 *physicalContext);
83
84 if (!commonPhysicalContextInterface->initialize())
85 {
86 lg2::error(
87 "Error initializing PhysicalContext Interface for Power Sensor for eid {EID} and sensor id {SID}",
88 "EID", eid, "SID", sensorId);
89 }
90 }
91 }
92
~NvidiaGpuPowerSensor()93 NvidiaGpuPowerSensor::~NvidiaGpuPowerSensor()
94 {
95 for (const auto& iface : thresholdInterfaces)
96 {
97 objectServer.remove_interface(iface);
98 }
99 objectServer.remove_interface(association);
100 objectServer.remove_interface(sensorInterface);
101 if (commonPhysicalContextInterface)
102 {
103 objectServer.remove_interface(commonPhysicalContextInterface);
104 }
105 }
106
checkThresholds()107 void NvidiaGpuPowerSensor::checkThresholds()
108 {
109 thresholds::checkThresholds(this);
110 }
111
processResponse(const std::error_code & ec,std::span<const uint8_t> buffer)112 void NvidiaGpuPowerSensor::processResponse(const std::error_code& ec,
113 std::span<const uint8_t> buffer)
114 {
115 if (ec)
116 {
117 lg2::error(
118 "Error updating Power Sensor for eid {EID} and sensor id {SID} : sending message over MCTP failed, rc={RC}",
119 "EID", eid, "SID", sensorId, "RC", ec.message());
120 return;
121 }
122
123 ocp::accelerator_management::CompletionCode cc{};
124 uint16_t reasonCode = 0;
125 uint32_t power = 0;
126
127 const int rc =
128 gpu::decodeGetPowerDrawResponse(buffer, cc, reasonCode, power);
129
130 if (rc != 0 || cc != ocp::accelerator_management::CompletionCode::SUCCESS)
131 {
132 lg2::error(
133 "Error updating Power Sensor eid {EID} and sensor id {SID} : decode failed, rc={RC}, cc={CC}, reasonCode={RESC}",
134 "EID", eid, "SID", sensorId, "RC", rc, "CC", cc, "RESC",
135 reasonCode);
136 return;
137 }
138
139 // Reading from the device is in milliwatts and unit set on the dbus
140 // is watts.
141 updateValue(power / 1000.0);
142 }
143
update()144 void NvidiaGpuPowerSensor::update()
145 {
146 const int rc = gpu::encodeGetPowerDrawRequest(
147 gpu::PlatformEnvironmentalCommands::GET_CURRENT_POWER_DRAW, 0, sensorId,
148 averagingInterval, request);
149
150 if (rc != 0)
151 {
152 lg2::error(
153 "Error updating Temperature Sensor for eid {EID} and sensor id {SID} : encode failed, rc={RC}",
154 "EID", eid, "SID", sensorId, "RC", rc);
155 }
156
157 mctpRequester.sendRecvMsg(
158 eid, request,
159 [weak{weak_from_this()}](const std::error_code& ec,
160 std::span<const uint8_t> buffer) {
161 std::shared_ptr<NvidiaGpuPowerSensor> self = weak.lock();
162 if (!self)
163 {
164 lg2::error("Invalid reference to NvidiaGpuPowerSensor");
165 return;
166 }
167 self->processResponse(ec, buffer);
168 });
169 }
170